HubSpot CTO on AEO: Why 'Citeworthy' Is the New SEO

How Dharmesh Shah Sees Content Strategy Changing for LLMs

Dharmesh Shah co-founded HubSpot on a simple thesis: millions of small businesses should benefit from the internet but found it intimidating. Now he's applying the same logic to AI. As CTO, he's both steering HubSpot's AI strategy and personally building agent.ai, a no-code agent platform with 50,000 agents created. This interview reveals how content strategy fundamentally changes when LLMs become your primary distribution channel.

On the new atomic unit: "Back then the atomic unit for the internet was web pages. That's what Google was indexing. Now the unit is actually smaller... it's the LLM looking at content, passing it through, synthesizing answers." The shift from pages to "chunks" changes everything about content strategy—you're not optimizing for clicks but for citations.

On being citeworthy: "In the same way that you need to be rankworthy, you need to be citeworthy... you're not trying to trick ChatGPT into providing your answer. Do the work to come up with really good answers." The philosophical continuity is striking: HubSpot's "be rankworthy" mantra from 2006 translates directly to "be citeworthy" in 2025.

On the translation metaphor: "Let's imagine you had 100 blog posts written in English for human readers... then you discover you have a bunch of people in Japan. You would translate to Japanese. What we're talking about now—from SEO to AEO—is translating for LLMs versus search engines." The mental model: treat LLMs as a new language your content needs to speak.

On snackable over narrative: "What's considered high quality is things that are easily quotable, snackable content by the LLM, that are pithy and precise... if you have a 400-word meandering answer, that's much less likely to be cited than something that provides exactly the answer in a nice sound-bitey form." Long-form narrative content optimized for human reading may underperform Q&A formatted content that LLMs can easily extract.

On agents as teammates: "Agents will effectively act like teammates. They may be junior teammates, they still need to be trained... in this hybrid world of carbon-based life forms and AI agents, how are you going to find these agents? There needs to be an equivalent of LinkedIn for agents." His vision for agent.ai: a professional network where agents have profiles, experience, and reputation.

6 Insights From Dharmesh Shah on Answer Engine Optimization

• AEO is the new SEO - Answer Engine Optimization solves for LLM citation instead of search ranking; the fundamentals (create value, don't game) transfer directly
• Reduce content to chunks - Break long-form content into question-answer pairs that LLMs can easily extract and cite; the atomic unit shrinks from pages to snippets
• Reputation compounds across eras - Content that ranked well in Google carries into Perplexity because early AI search used traditional search as input; investments in quality compound
• Negative returns are real - Building a reputation as "AI slop" or low-quality content is worse than zero returns; LLMs will remember and deprioritize you
• Track AI traffic separately - Your organic traffic is likely down because attention moved to answer engines; create a distinct bucket for ChatGPT/Perplexity referrals
• Agents need professional networks - Agent.ai's thesis: 50,000 agents exist, they need discovery and reputation systems just like human professionals

What This Means for Content and Marketing Teams

Dharmesh's framework reframes AI content strategy as a translation problem: same ideas, new language. The key insight for businesses isn't that SEO dies—it's that you now have two audiences: humans who read and LLMs who cite. Content optimized for human consumption may underperform content optimized for LLM extraction. Organizations that understood inbound marketing already have the philosophical foundation ("create value, don't game"); now they need the tactical adaptation (chunks over narratives, Q&A over prose, citeworthy over rankworthy).